Development of a citizen participation public service innovation model based on smart governance

Abstract

This study explores an efficient approach to providing customized public services through a smart governance-based public service innovation model (SG-PSIM) that combines intelligent technology and co-creation. Multiple methodological approaches are applied to develop and evaluate the proposed SG-PSIM. Intelligent methodologies that can support the public service policy process are discussed, and the applicability of the SG-PSIM is demonstrated through four case studies. The study results showed that SG-PSIM can effectively collect the opinions of citizens in diverse ways and provide opportunities for citizens to actively participate in the development of public service policies.

Introduction

The current era of the Fourth Industrial Revolution is represented by intelligent technologies such as the Internet of Things (IoT), artificial intelligence (AI), and big data (Oosthuizen 2022). World-class organizations, such as Apple and Google, create new business opportunities by utilizing these technologies that make our lives convenient. For instance, fresh food is delivered early in the morning through Amazon without transferring funds through mobile banking or physically visiting stores. In addition, in the investment sector, daily lives are rapidly shifting to digital, where traders are robot advisors. This phenomenon is referred to as digital transformation. Vial (2019) defines digital transformation as “a process that aims to improve an entity by triggering significant changes to its properties through combinations of information, computing, communication, and connectivity technologies.” In other words, it is a disruptive innovation that rapidly improves traditional operational methods and services by utilizing intelligent technologies. As intelligent technologies spread throughout society, they are expected to have a significant impact on government operations as well.

Hong et al. (2017) proposed the concept of “smart governance,” which pursues the active participation of citizens and stakeholders by applying intelligent technology and the concept of co-creation. They defined smart governance as a new government innovation system that combines intelligent technologies to ensure the participation of various stakeholders throughout the entire government policy-making process and overcome the constraints of existing governance. The evolution to smart governance causes significant changes in the public service sector. For example, Eom and Lee (2022) pointed out that the new intelligent technologies made possible for governments “provide citizens with more tailored public services, forecast with greater accuracy, and simulate complex systems ranging from military operations to the private sector of entire countries.” Ryu et al. (2022) also addressed that public services need to reflect the opinions and needs of the general public. In addition, they emphasized that introducing intelligent technology is necessary to avoid any distortion of citizen opinions due to limited survey samples or arbitrary survey methods, even when data collection is done through commercial polling agencies. Thus, governments worldwide will need to actively maximize the use of intelligent technology to provide convenience to and enhance the participation of citizens who are beneficiaries of public services. In the past, governments focused on building information and communication infrastructure and systems to provide public services (Hong et al. 2017). However, consumer-centered public services based on these infrastructures, which citizens can easily access and benefit from, will be required in the future. Sarker et al. (2018) argued that applying big data in the public sector could ensure fast delivery, high efficiency, accuracy, and transparency of public services. New opportunities for using intelligent technology in public service delivery related to AI, big data, machine learning, and service automation have been discussed as necessary for public authorities (Lindgren et al. 2019). Therefore, digital government transformation and smart or intelligent governance can enable the delivery of consumer-focused public services.

Efficient and effective public services that meet citizens’ needs should be provided on time (Araujo et al. 2013; Hartley 2023). Thus, public service policy development should be preceded by collecting opinions from service users and reflecting them in policies. However, the current public service process may need more direct communication with citizens and an active collection of opinions (Lindgren et al. 2019). In addition, several questions have been raised about whether methods and techniques have been established to analyze citizens’ opinions on public services objectively (Scholta et al. 2019; Kowalski et al. 2020). To provide citizens-oriented public services through digital infrastructure, it is first necessary to understand how digital technologies affect the interaction between citizens and policymakers in public service provision (Lindgren et al. 2019).

As digital transformation or infrastructure is introduced into various industries and daily lives, research on new ways of participation in which citizens can freely participate using intelligent technologies and reflect their opinions is required. The COVID-19 pandemic has dramatically changed the way we work with the required social distancing measures and travel restrictions (Lee and Lee 2021). Thus, the pandemic crisis reinforced the need for innovation, especially agile and resilient delivery of public services. In particular, the pandemic has accelerated the use of advanced information and communication technologies (ICTs) to facilitate various innovative contact free service delivery models (Lee and Lee 2021). During the pandemic, governments became agile organizations supported by advanced ICTs for public service delivery with citizen collaboration and participation (Ryu et al. 2022; Hartley 2023). The COVID-19 pandemic demonstrated that citizens' participation is essential for the implementation of the smart governance model (Hong et al. 2020; Ryu et al. 2022; Hartley 2023). However, there is a paucity of practical, academic, and/or empirical research. Therefore, this study aims to suggest the possibility of providing consumer-centered public services by applying smart governance that combines intelligent technology and co-creation between government and citizens. To provide customized public services, we suggested a smart governance-based public service innovation model (SG-PSIM), which pursues the intelligentization of four stages of public service policy processes: establishing agendas, formation, decision and execution, and evaluation.

For the development of SG-PSIM, the following three step-approach is applied as follows. First, a thorough review of the literature on the public service policy process, smart governance, and the text mining analysis procedure was performed. Second, the limitations of the existing method were examined for each policy stage, and intelligent models that can be used throughout the public service policy process based on citizens’ opinions were proposed. Third, four case studies were analyzed for each policy stage to test the new model’s applicability. This study is novel in that it presents a generic model that ensures citizen participation in the entire policy stages through theoretical consideration and is expected to overcome the limitations of the existing method.

The rest of the paper is organized as follows. The following section introduces the concept of smart governance and the intelligent technologies to implement it. Section 3 presents a research design and intelligent models for each policy stage. In Sect. 4, four cases at the stage of the public service policy process are presented to explore the applicability of SG-PSIM, and summarizes SG-PSIM. Conclusions, contributions, and limitations are discussed in Sect. 5.

Literature review

Public service process

General public services comprise a four-stage process: establishing policy agendas, policy formation, policy decision and execution, policy evaluation, and feedback (Shiffman 2008). In the first stage, establishing policy agendas refers to a government act converting social issues into policy questions, which politicians should focus on, to solve them officially. In other words, a social problem that impacts the citizens is initially identified (McCombs 2005).

In the policy formation stage, policy alternatives are developed by exploring and discovering policy methods. When the search and development of policy alternatives are completed, priorities are determined by applying criteria that compare and evaluate policy alternatives. As such, policy formation develops specific alternatives to achieve policy goals and reviews existing alternatives to modify, supplement, or develop new alternatives, according to the situation (Howlett and Ramesh 2003). Subsequently, a comparative and evaluative method for measuring the efficacy of each alternative policy is offered.

During the third stage, in the policy decision and execution stage, decision-makers formulate policy proposals and implement public budget execution based on the policy alternative proposed through the policy formation process. Specifically, decision-makers determine the specific course of action that will best solve most citizens’ problems and put the chosen option into action (Schofield 2001).

In the last stage, policy evaluation is an overall activity that evaluates whether policy alternatives have been implemented as planned and whether policy goals have been achieved. It presents the cause of policy success and failure, along with securing responsibility in the policy process (Drissa et al. 2019). In addition, policy evaluation is conducted to improve the efficiency and effectiveness of service operations. Finally, feedback is provided based on the results of the evaluation. The public service policy process is shown in Fig. 1.

Fig. 1.

The public service policy process

Smart governance: combining co-creation and intelligent technology

In recent times, citizens can easily propose their ideas and suggestions for public services and participate in policy evaluation reviews using mobile platform devices like smartphones. These technologies enable cooperative communication between citizens and the government previously limited by time and cost constraints. Therefore, it is necessary to establish an intelligent system where citizens express their opinions, and the government creates values based on their voices. In other words, the four-stage public service policy process, from gathering citizens’ needs to citizens’ policy evaluation, must be intelligentized. For this purpose, the concepts of intelligent technology, digital transformation, and smart governance related to the model we are developing are explained.

Advancement of intelligent technology

A key feature of intelligent technologies is their ability to self-learn and adapt to new situations based on the data obtained (Marinov 2019). These technologies include AI, big data, and text mining. Using these, the government discovers customized public services centered on citizens and changes the delivery process of existing public services more efficiently and effectively.

1. With the Fourth Industrial Revolution, the AI era is approaching. AI is profoundly penetrating both the private and public service sectors. Countries worldwide are shifting to AI-based intelligent governments by implementing digital, ubiquitous, mobile, and smart governments (Kankanhalli et al. 2019). For instance, the city of Boston, Massachusetts, USA, collects and analyzes civil complaints using Boston 311 Call, “an easy-to-remember telephone number that connects you with highly-trained constituent service representatives” (City of Boston) and an exclusive app to solve road damage, graffiti, garbage disposal, and parking problems. Nevada health authorities have shifted restaurant health inspections from random to target selection using AI (Sadilek et al. 2018). As a result, AI offers optimal and efficient tailored services, which are expected to be widely utilized in the future.

2. Big data is mainly considered a means of improving the competitiveness of companies; however, application in the public service sector has been increasing (Munne 2016). For instance, Jeonju City in South Korea analyzed boarding and depart information of the public transportation system based on time-specific passenger transfer information using traffic card data and floating population information using communication data. Ensuingly, the city decided on more accurate and reasonable public transportation routes and new route decisions (Ministry of Land, Infrastructure, and Transport of Korea 2018). London, England, imposed congestion charges based on traffic data to reduce chronic traffic congestion, which resulted in increased public transportation use, improved services, and improved atmospheric conditions (Givoni 2012). Big data can also be used to provide better citizen safety in smart cities where different types of data are collected, including sensors, social media, and emergency calls (Munne 2016).

3. Social media, a tool to share individual opinions and experiences, is also used in the public service sector (Kowalski et al. 2020). Text mining is a technique for analyzing text data containing vast personal opinions on social media. There are many advantages to analyzing big text data. First, by using techniques such as the classification of unstructured data and sentiment analysis, it is possible to derive meaningful information hidden in numerous texts (Iqbal et al. 2022). Second, it is possible to make faster decisions by finding specific patterns or trends (Hassani et al. 2020). Third, various data sources provide insight into social problems (Hassani et al. 2020). Fourth, extensive text-data analysis in the public sector enables ‘evidence-based policy’ promotion to establish and implement policies (Bright and Margetts 2016; Androutsopoulou and Charalabidis 2018). In the future, as text mining directly analyzes citizens’ opinions through online platforms such as individual comments, electronic civil petition bulletin boards, and newspaper articles, it is expected that government policymakers’ interest and utilization will increase.

Digital transformation

Innovation is generally divided into progressive and radical innovation depending on speed, and digital transformation emphasizes the occurrence of rapid changes. In addition, digital transformation does not change company environments using information technologies (Chin et al. 2023). However, it refers to changes in paradigms applied to all fields, from product development to production and marketing. Platform companies, such as Google, Microsoft, Amazon, and Alibaba, provide customized services to consumers by building and operating intelligent technology-based platforms (Rha and Lee 2022; Chin et al. 2023).

Digital transformation in daily life can be found easily (Chin et al. 2023). In the past decade, advertising revenue for TV and newspapers has nearly halved, whereas Web Portal’s and YouTube’s market share exploded. In retail services, sales at hypermarkets and department stores decreased, whereas online sales increased rapidly. One-third of U.S. department stores and even more small stores have closed. In the financial industry, 65.3 percent of the U.S. population uses digital banking services as of 2022 (Statista 2023). These changes are also taking place in government policy or public services. As Rha and Lee (2022) pointed out, digital transformation in the service sector is emerging as an essential study domain. For instance, the “Tell Us Once” program run by the United Kingdom provides a one-stop service; “notifications to tax authorities, the passport office, local governments, and benefits programs” (Cornwall Council, United Kingdom 2023).

Smart governance

The establishment and operation of national policies have changed from top-down governance to joint participation between the government and the private sector (Emerson et al. 2012). Unlike government-centered state administration, governance refers to state administration in which stakeholders solve problems based on shared interests and opinions without unilateral government initiatives (Bryson et al. 2006). The previous government system was suitable for uniform societies, where most members faced similar problems. Conversely, governance began with the recognition that traditional bureaucratic methods were inappropriate in various situations, such as decentralization, privatization, marketization, and globalization in national administration (Grindle 2004). This emphasizes the importance of participation and cooperation for the common goal of sustainable social development, in which the government sees the people as partners.

Recently, governance centered on cooperation has resolved social and political issues. Governance, which emphasizes cooperation with the private sector, away from government-led policy implementation, is achieving results in creating and implementing policies through collaboration among the government, citizens, and stakeholders (Emerson et al. 2012).

To promote the awareness and positive perceptions of smart operational policies among the general public, the government needs to incorporate intelligent technologies into their communication and engagement strategies when announcing the features of intelligent projects (Hartley 2023). Strategic communication efforts are needed to realize innovative initiatives in collaboration with citizens. Intelligent technologies should enable easy citizen participation that can benefit both citizens and the government (Arku et al. 2022; Cole et al. 2023). Specifically, for the effective implementation of smart government policies, the general public should have convenient access to high-quality information, multidimensional media profiles, and clear messages (Hartley 2023).

However, in the existing governance structure, due to physical and time constraints, the policy establishment and implementation through agents rather than the direct participation of stakeholders were inevitable. In modern society, where individual needs are diversified and expressing opinions is easy, cooperation and participation with a wide range of stakeholders are required, which differs from the existing governance system with limited stakeholder participation. As such, the development of intelligent technologies provided conditions for stakeholders to participate easily and quickly, enabling the shift toward smart governance that incorporates intelligent technology into existing governance.

Hong et al. (2017) proposed the concept of smart governance. It is a conceptual innovation system that combines information, communication, and intelligence technology with the concept of co-creation to ensure the participation of various stakeholders in establishing policy and execution and to overcome the constraints of existing governance. Therefore, the core value of smart governance is to provide objective and efficient policies or demand-based public services by analyzing citizen opinions with intelligent technology.

Research design and intelligent models

Research design

A three-step approach was applied to conceptualize and test the applicability of a new model (see Fig. 2). First, considering that policy pertaining to big data consists primarily of text, the public service policy process was defined, and the text mining approach was described. Second, the limitations of the existing method were examined for each policy stage, and an intelligent methodology that could be used throughout the public service policy process based on citizens’ opinions was proposed. Third, to examine the applicability of our proposed model, four case studies were reviewed with implications for each policy stage, and then finally, SG-PSIM was proposed.

Fig. 2.

Research design

Text mining as a tool for analyzing citizen opinion

In intelligent societies, the interest of citizens in public services is growing. The government provides public services to people from various classes; however, in many cases, policies need to be adequately promoted and be suitable for user needs (Drissa et al. 2019). Therefore, for public service policies to succeed, it is important to establish policies to understand users’ interests, requirements, and evaluations and look for improvements.

Public service analysis is generally based on economic effects and survey results that appear as quantitative data. However, to realize society’s values, it is necessary to understand the process of identifying universal values among citizens and expressing collective intelligence (Kenter et al. 2015). Therefore, citizens’ collective intelligence results should be reflected in the policy process. The expression of collective intelligence strengthens the right to speak to the government by most citizens. In addition, improved communication and opinion collection will strengthen the efficiency of public service policies and the ability to respond to crises. However, as the opinions of various stakeholders should be collected, and the criteria for policy decisions should be presented to stakeholders, a machine algorithm that can process unstructured big text data and imitate the expression process of collective intelligence is required. Text mining is rapidly developing as one of the machine learning technologies that can quantify unstructured data.

This model is based on the text mining which undergoes a visualization process after data collection, preprocessing, and analysis. Firstly, public opinions are collected using SNS and conventional media (Bright and Margetts 2016). The data collection utilizes such automated methods as web crawling, web scraping, and open application programming interface (API). Before data are analyzed by applying various methodologies to identify issues and user opinions, data are preprocessed. In the preprocessing stage, morphological analysis is required to determine the structure of various linguistic properties such as parts of speech, suffixes, and prefixes (Ritchey 2006). Data preprocessing involves data cleaning, data integration, data conversion and transformation, and data reduction.

With data cleaned, data analysis is performed based on the research purpose. The most popular analysis methods include clustering, social network, classification, sentiment, and facet analyses. Lastly, the results are visualized to allow users and public institutions to understand the information easily.

Intelligent model for establishing policy agendas

Traditional methods for establishing policy agendas include Delphi, literature reviews, and surveys. Existing methodologies mainly focus on direct stakeholders; consequently, the general public lacks opportunities to express individual opinions. For instance, surveys examining public opinions may have a limited sample, potentially causing bias. Next, issue triggers are detected through issue raisers, broadcasts, newspapers, briefing sessions, offline meetings, and online data collection, and various public opinions are derived and reflected in policies through big data analysis. Figure 3 shows the difference between the existing methodology and the new model.

Fig. 3.

Changes in establishing policy agendas

source: Hong et al. (2020), p. 120

The intelligentized method presented in Fig. 4 can be used to establish a policy agenda that overcomes the limitations and disadvantages of the existing practice. As the first step for establishing an intelligentized policy agenda, social issues are detected through conventional and social media monitoring. Next, issue triggers are detected through issue raisers, broadcasts, newspapers, briefing sessions, offline meetings, and online data collection, and various public opinions are derived through big data analysis and reflected in the policy. Finally, the agenda derived through visualization of analysis results is intuitively shown, and the final agenda is determined.

Fig. 4.

Intelligentizing stage of establishing policy agendas

source: Hong et al. (2020), p. 85

Intelligent model for policy formation

Traditionally, small elite groups with expertise in related fields have selected the optimal alternative in policy formation. However, as society becomes complex and rapidly changing, it is becoming difficult for a small number of experts to decide on reasonable policy alternatives accordingly. Therefore, it is necessary to form policy using collective intelligence, such as public deliberation. However, identifying active citizens, collecting opinions, mediating conflicts, and reaching a consensus take time and cost. Intelligence techniques that support stakeholder engagement, information sharing, and consensus building overcome these limitations.

Figure 5 shows the difference between the existing methodology and the new model. Policy formation in the new model utilizes intelligent technology, as shown in Fig. 6. First, using the intelligent platform, ideas presented by stakeholders in text form are collected and preprocessed. Second, in the process of collecting policy alternatives, the presented ideas are categorized into similar areas and visualized using clustering and automatic classification technology to evaluate ideas. In addition, deep learning can be utilized to prioritize alternatives.

Fig. 5.

Changes in the policy formation process

source: Hong et al. (2020), p. 117

Fig. 6.

Intelligentization stage of policy formation

source: Hong et al. (2020), p. 120

Intelligent model for policy decision and execution

As the existing policy decision process has relied on decision-making by policymakers, policy decisions might not be optimal due to differences in opinions between citizens and policymakers. Figure 7 shows the limitations of the existing methodology and differences in the new model. The new model proposes a method for intelligentizing the trust formation process and execution in the existing policy decision and execution process. Figure 8 shows the process of forming the legitimacy of these policy decisions and intelligent technology that allows citizens to monitor policy execution.

Fig. 7.

Changes in policy decision and execution processes

source: Hong et al. (2020), p. 52

Fig. 8.

Intelligentization stage of policy decision and execution

source: Hong et al. (2020), p. 153

Intelligent model for policy evaluation

Existing policy evaluations, such as questionnaires, interviews, and observations, are complex and require time and money. In addition, policy evaluations that rely on the experience or intuition of experts may be subjective. Recently, with the development of opinion mining that quantifies unstructured data, text analysis of SNS and newspaper articles has become possible. Therefore, the intelligent model presents a method of evaluating policies by analyzing citizens’ thoughts and opinions, as shown in Fig. 9. Applying this model to policies can provide more efficient and objective evaluations, positively impacting public service in the future.

Fig. 9.

Changes in the policy evaluation process

source: Hong et al. (2020), p. 183

As shown in Fig. 10, sentiment analysis and topic modeling are mainly used for policy evaluation. Sentiment analysis is employed to determine whether citizens’ favorable and unfavorable ratios for policies and TF-IDF(Term Frequency-Inverse Document Frequency) is appropriate to derive the positive and negative factors. Recently, machine learning, faster and more accurate than dictionary-based sentiment analysis used in the early days, is mainly applied.

Fig. 10.

Intelligentization stage of policy evaluation

source: Hong et al. (2020), p. 185

Exploring the applicability of intelligent models through case studies

To conduct an in-depth analysis of the applicability of a proposed model to public services, we examine the operational procedures and methods implemented in several cases. The following four cases were investigated based on the information released with the citizen participation strategy and experience of participation directly or indirectly in these case studies.

Case study 1: Analysis of complaint trends using topic modeling

This case study aims to analyze citizens’ complaints posted on the electronic complaint board to identify which problems are continuously occurring or decreasing; thus, it corresponds to the intelligent model for establishing policy agendas. This case is summarized and reorganized based on the research of Park and Lee (2020).

Research procedures and methods

The research procedure is shown in Fig. 11. First, 9,625 electronic complaints have been collected over three years (2015–2017) to identify trends of complaints raised through Busan’s electronic bulletin board. Subsequently, only nouns were extracted through the preprocessing process, and 20 civil complaint issues were identified using the Latent Dirichlet Allocation (LDA) topic modeling, which is known to be suitable for clustering. Individual topics that showed an upward trend based on the regression analysis were classified as hot complaints, whereas those with a downward trend were classified as cold complaints.

Fig. 11.

Research procedures and methods

source: Park and Lee (2020), p. 71

Data analysis results

Out of 20 topics, 6 are traffic complaints, and 14 are non-traffic related complaints, such as festival complaints and complaints regarding welfare for the elderly. In addition, an analysis of the trend in traffic complaints, which accounted for the most significant proportion, between 2015 and 2017 reveals that most complaints are regarding bus stops, bus allocation intervals, and bus lanes. Furthermore, the regression analysis identifies four very upset complaints about bus stops and taxi drivers and four callous complaints regarding CCTV installation and bus routes.

Implications for establishing policy agendas

Complaints revealed unfriendly drivers, buses that did not stop, non-compliance with arrival times, denial of passengers, boarding or departure of passengers, and violation of safe driving practices. This suggests that the city of Busan in Korea needs strict administrative measures to improve customer response training and guidance for bus and taxi drivers and prevent violations. In addition, such callous complaints as CCTV installation, bus routes, park facilities, and festivals suggest the need to monitor whether the issues have improved or been temporarily reduced.

Case study 2: Automatic classification of traffic complaints in Boston

This case is a study on the automatic classification of civil complaints in Boston, USA, and corresponds to an intelligent model in policy formation. In this stage, the process of collecting opinions should be intelligent, such as aggregating ideas of experts or stakeholders and summarizing and presenting the results. Boston presents a case of an intelligentized method of automatically classifying traffic complaints. This case is summarized and restructured based on the research of Kim and Hong (2021).

Research procedures and methods

This study aims to identify the optimized category and automatically classify traffic complaints and employ the CNN (Convolutional Neural Network) model, which can be used for data classification, with complaints categorized into those that occurred from January 2016 until November 2018 in the City of Boston, USA. The deep learning-based clustering and automatic classification process are shown in Fig. 12.

Fig. 12.

Research methods

source: Kim and Hong (2021), p. 4

Data analysis results

First, the CNN model was used with data on 5,000 civil complaints from 2016 to 2017, and an overfitting problem occurred. Overfitting is a problem caused by overlearning the features of the training data set. Therefore, in the second stage, the CNN model was applied to 9,000 complaints, including those that occurred until 2018. However, the accuracy was lower than that of the existing model; therefore, optimization for the existing category was conducted. In the third stage, the optimized category was determined using the clustering technique and CNN; however, accuracy could not have improved significantly. Therefore, in the fourth stage, the Synthetic Minority Over-Sampling Techniques + Edited Nearest Neighbors (SMOTE + ENN) were used to solve the problem of data imbalance, in which bias occurs in a specific category. This produced a high level of accuracy at 85%–90%.

Implications for policy formation

As the participation of both citizens and experts in the policy formation process is important, automatic classification techniques are essential to disseminate and converge ideas from various stakeholders intelligently. The results of this study have implications in that it proposes a technique for automatically classifying big data of discussions on various policy alternatives conducted on policy formation platforms, allowing dissemination and convergence. Using such an intelligent model of policy formation produces quality improvement in discussions for policy formation and reflects the discussion results in public service policy.

Case study 3: automatic analysis of minutes using the facet learning model

Lee et al. (2020) analyzed public discussion using an intelligent model during the policy decision and execution stage. This study proposes a method for citizens to easily understand the public discussion process, such as Congress minutes, which are voluminous and difficult to understand, by focusing on specific attributes. This research corresponds to the intelligent model for the policy decision and execution stage. This case is summarized and reorganized based on the research of Lee et al. (2020)

Research procedures and methods

This study employed a machine learning algorithm with 298 min of the seventh Busanjin-gu Basic Council in Busan, Korea, collected from the Council website. Specific properties were divided into three facets, as shown in Table 1. Facet A was to understand the purpose of the speech, Facet B was to identify whether the speech evidence was presented and the characteristics of the evidence, and Facet C was to understand the cause of the conflict.

Table 1.

Facet classification

	Definition	Classification of facets
		1	2	3	4
Facet A	Purpose of statement	Social function	Administrative function	Economic function	Statements that do not fall under 1, 2, 3; Statements related to procedure; Greetings
Facet B	Presentation of evidence and characteristics of evidence	Measurable evidence, such as statistical facts, scientific facts, and social commitments	Unmeasurable evidence, such as psychological facts, sensory concepts, and ethical facts	No evidence presented
Facet C	Cause of conflict	Conflict of principle	Conflict of rules	No conflict

Lee et al. (2020), p. 54

Data analysis results

They generated a training set for machine learning based on a pre-prepared facet analysis framework to quantify properties. First, 33 of the 298 total minutes were randomly extracted. A total of 7,475 remarks were recorded, and three researchers examined the context of the discussion process and carefully read the contents to classify the nature of the remarks according to the facet framework. The recorded remarks were divided into learning data (5,741 remarks) and test data (1,734). Subsequently, machine learning was conducted using a learning model based on the Long Short-Term Memory (LSTM) algorithm.

Facet A analysis revealed that 49% of the remarks were on social, 9% on administrative, and 21% on economic functions. Facet B indicated that there were twice as many remarks based on unmeasurable (24%) than measurable evidence (11%). The results of Facet C showed that 73% of the remarks did not contain controversial content, 4% of controversial remarks were fundamental conflicts, and 23% were conflicts over rules.

Implications for policy decision and execution

Facet analysis drastically reduces the amount of information for citizens interested in reviewing minutes by automatically classifying points of order and expressions of intent. In addition, with 62% accuracy for more than 100,000 statements, it was possible to identify the nature of parliamentary statements, the presence or absence of arguments and evidence, the classification of controversial statements, and whether the cause was by principle or rule. Moreover, public discussions on various aspects can be analyzed using a facet analysis framework for specific speech attributes required by researchers or citizens to improve policy governance.

Case study 4: Sentiment analysis of visitors to Gamcheon Culture Village

Hong et al. (2019) investigated an intelligent model of the policy evaluation stage by identifying citizens’ opinions on a specific policy after its implementation and demonstrated the process of analyzing tourists’ positive and negative opinions, including reasons, based on visitor reviews of Gamcheon Culture Village using sentiment analysis. This case is summarized and restructured based on the research of Hong et al. (2019).

Research procedures and methods

The case of Gamcheon Culture Village, a representative tourist attraction in Busan, is considered a successful model for urban regeneration projects that combine cultural contents. A total of 2,113 reviews from Google regional data were collected to evaluate the performance of urban regeneration projects from the perspective of visitors in this village, and positive and negative factors were derived using the Term Frequency-Inverse Document Frequency (TF-IDF) weight model.

Data analysis results

The sentiment analysis revealed that approximately 80% of visitors expressed positive opinions, indicating high satisfaction (refer to Table 2). In addition, keywords extracted from positive reviews demonstrated that “photo” was used together with keywords such as “the little prince,” “mural,” and “attractions,” indicating that these contents reminded visitors of Gamcheon Culture Village. Negative reviews indicated that the keyword “people” was used in reviews stating, “There are too many people to tour around,” as well as sentences with the keywords “little prince,” “photo,” “waiting line,” and “time.” Reviews reported that taking a picture with the Little Prince sculpture took a long time, as the line was too long.

Table 2.

Sentiment analysis results

Sentiment	Positive	Negative	Total
Number of cases	1694	419	2113
%	80.17%	19.83%	100%

Hong et al. (2019), p. 1239

Implications for policy evaluation

Sentiment and TF-IDF analyses can be used in the policy evaluation stage. The highly positive evaluation of Gamcheon Culture Village indicated that the urban regeneration project run by Busan, Korea was successfully underway. In addition, the machine learning method can be employed to determine policy consumers’ positive and negative perspectives from a vast amount of text. Moreover, the results indicated that these analyses could be implemented for future policy evaluation, as they provide a method for automatically classifying positive and negative factors through TF-IDF analysis.

A summary of the four cases discussed above is shown in Table 3. The four case studies demonstrated the applicability of our model to the real-world public service policy process. They showed how the new model could facilitate a better understanding of policymakers on citizens’ opinions.

Table 3.

Case summary

Policy stage	Case	Research method	Implications
Establishing policy agendas	Case 1: Analysis of complaint trends	Topic modeling	A total of 20 issues necessary for policy reflection were derived through complaint analysis, along with the identification of four hot issues with constant complaints
Policy formation	Case 2: Automatic classification of traffic complaints in Boston, USA	Clustering based on deep learning and network analysis	The automatic classification technique provided a basis for disclosing and collecting various policy proposals (big data for discussion) made in real-time on the policy formation platform in the future
Policy decision and execution	Case 3: Automated analysis of minutes	Facet learning model	Established a foundation for citizens to easily identify the complex process of public discussions, centered on specific attributes
Policy evaluation	Case 4: Sentiment analysis of Gamcheon Culture Village’s visitors	Sentiment analysis and TF-IDF	Established the basis for evaluating public policies, by identifying the positive/negative rate of tourists visiting Gamcheon Culture Village, and automatically analyzing the positive/negative factors

Smart governance-based public service innovation model

This research aimed to propose an SG-PSIM to explore the possibility of providing customized public services. Based on the study of Hong et al. (2020), the SG-PSIM was finally proposed through a three-step (reviewed previous studies—presented SG-PSIM—and confirmed using existing cases) approach. Table 4 summarizes the existing methodology, a flow of SG-PSIM, and related cases with implications for SG-PSIM. This model emphasizes the convergence of citizens’ opinions from establishing the policy agenda stage to the final policy evaluation stage. Accordingly, the optimal method, tools, and intelligent technology that can reflect citizens’ opinions at each stage were presented. The SG-PSIM will be able to collect the opinions of the digital and familiar generation in more diverse ways and provide an opportunity for citizens to participate in public service policies actively.

Table 4.

Summary of SG-PSIM with cases

Policy stage	Existing methodology	SG-PSIM	Case	Research method	Implications
Establishing policy agendas	%1 Delphi technique %1 Literature research %1 Survey	Intelligent model for establishing policy agendas based on limitations and shortcomings of existing practices (see Figs. 3 and 4)	Case 1: Analyzing complaint trends	Topic modeling	Topic modeling can be used to identify the opinions or demands of a large number of citizens in text form by issue, suggesting that intelligent technology can be of good use in setting policy agendas
Policy Formation	%1 Public discussion %1 Town hall meeting %1 Citizen jury	Intelligent model for policy formation based on innovative technology (see Figs. 5 and 6)	Case 2: Automatic classification of traffic complaints in Boston	Clustering based on deep learning and network analysis	In the policy formation stage, policy proposals (big data for discussions) from various stakeholders should be released and gathered in real-time. For this process, the use of the automatic classification is suggested
Policy decision and execution	%1 Macro discussion %1 Presenting data based on policy decisions through minutes %1 Monitoring/supervision of policy implementation	Intelligent model for policy decision and execution based on citizen monitoring of policy implementation (see Figs. 7 and 8)	Case 3: Automated analysis of minutes	Facet learning model	To make policy decisions, stakeholders need to understand the content of policy discussions. Suggestively, facet analysis enables the automatic identification of specific attributes
Policy evaluation	%1 Observation %1 Literature review %1 Survey/interview	Intelligent model for policy evaluation based on citizens' thoughts and opinions (See Figs. 9 and 10)	Case 4: Sentiment analysis of Gamcheon Culture Village’s visitors	Sentiment analysis and TF-IDF	Suggestively, policies can be evaluated through sentiment analysis, and positive and negative factors for policies can be derived through TF-IDF

The public service innovation model based on smart governance can provide citizens and public officials with various new value creating opportunities. First, for citizens, they can easily express their opinions based on intelligent technologies and have ubiquitous access to relevant government information. Second, they have opportunities to reduce time and costs in conveying their opinions and concerns. Third, as diverse opinions can be converged quickly by the model, citizens would most likely receive the various services they desired.

From the government's perspective, first, with citizens’ active participation and the collection of diverse opinions, this model can help with more objective public service delivery. Second, the SG-PSIM can provide transparency in decision-making by collecting citizens' opinions using intelligent technologies. In addition, SG-PSIM can gain citizens' trust in government agencies and achieve cost savings compared to past opinion polls to support decision-making.

With the support of digital technology and data-based decision-making, the government can provide better services that reflect citizens' needs, thus improving efficiency, effectiveness, and transparency in public service delivery. Therefore, the SG-PSIM proposed in this study can positively benefit both citizens and the government.

Discussion and conclusions

Along with the recent wave of the Fourth Industrial Revolution, digital transformation is leading our society to transform into an intelligent society by disseminating intelligent technology in most areas, including private sectors and government. In an intelligent society, all objects, such as machines to machines and machines to humans, will be connected to facilitate data collection and information exchange. Personalized services can be provided using big data and AI analysis capabilities. In particular, public services are government activities to solve the problems of members of society and provide a more desirable environment according to members’ needs. Understanding the needs and participation of members of society is of paramount importance. To this end, the government should provide platforms and systems to allow citizens to participate actively and provide opinions in public services. In addition, it is necessary to accurately analyze the opinions of members of society on public services to develop them into customized public services that all members can benefit from. Therefore, this study presented SG-PDIM applying the concept of smart governance and intelligent technology. Moreover, by analyzing how SG-PSIM, presented through four examples, can be applied to public services, we propose the applicability of this model. The case studies indicated that the digital footprints of individual citizens could be tracked to identify policy issues they require and evaluate government policies objectively through text mining. Therefore, our model can help practitioners and academics to plan and implement tailored public service based on citizen opinion.

To achieve the goal of SG-PSIM, the model requires a nation-wide infrastructure. However, the process of collecting and analyzing data from the general public through the national infrastructure may result in data leakage. This challenge may give rise to third-party issues (e.g., crime), and lead to questions regarding data reliability. Although SG-PSIM can improve decision-making based on citizen participation for the public interest, it should be implemented while protecting privacy, data security, and transparency. In addition, the digital divide issue should be managed from the same perspective as the economic divide, addressing it through active government support. One of the solutions could be to create easily accessible and user-friendly platforms that are ambient to citizens. In addition, providing opportunities for education and training on digital device usage could also be a way to reduce the digital divide.

Implications of the study results

This study has academic significance in that it considers that citizen participation can be guaranteed by applying intelligent technology at each stage of the public service policy process and presents SG-PSIM as suitable for the public service field based on real-world cases. SG-PSIM has the advantage of being able to support efficient analysis of diverse and vast civic opinions through rapid and systematic procedures by applying intelligent technologies and smart governance. Governments can use this model to provide consumer-centered public services. In addition, big data analysis will allow the identification of the characteristics of various citizens’ opinions, prediction of public service demand, and provide citizens with preemptive convenience, which will reduce the possibility of additional complaints while increasing the efficiency of public service work.

The practical implications of this study are as follows. First, this model can be employed by public service policymakers. It is possible to collect meaningful opinions and ideas using the AI algorithm required to collect opinions from various stakeholders on the platform and seek ways to implement their opinions. Second, this model can be used not only in the public field but also in other service fields that require processing big-text data. For instance, platform-based service organizations can use this model to identify, analyze, and utilize customer needs. Third, this model can be used as basic analysis process materials for collecting opinions presented in structured or unstructured data from members of society or specific customers and making policy decisions. For instance, using media content-based data, customer requirements, and participation levels can be reviewed to analyze competitiveness and strengthen the customer experience. Lastly, the COVID-19 pandemic has prompted numerous new innovations in various organizations and industries. The fear of contacting the virus encouraged consumers to avoid face-to-face encounters with service providers. Thus, “untact” service methods based on advanced ICTs became prevalent, such as online conferences, Teladoc, and Zoom meetings (Lee and Lee 2020; Chin et al. 2023). In the public service area, a smart governance model can increase citizen participation with contact-free encounters to co-create value, identify citizens’ needs, and also explore new solutions to existing issues. Also, it enables active engagement of the various stakeholders while minimizing exposure risks to the virus (Arku et al. 2022). Furthermore, as the current generation citizens are quite familiar with digital devices, such as smartphones and wearable devices, their increasing participation is plausible for implementing the smart governance model.

Limitations and future research directions

The limitations of this study and future research areas are as follows. First, further research on intelligentization in other processes that can minimize human intervention in the policy process is required since this research showed that partial processes at each stage can be intelligentized. Second, the innovation model proposed in this study focused on the technical part. Given that the successful implementation of this model can be influenced by not only technical factors but also organizational culture and individual competence or interest, research on these questions is required. Third, the direct participation of citizens in the policy process will change the identity of members of society. Research in this field is necessary, as it is expected to significantly impact the role of public opinion formation of the two pillars, policy expert groups and mass media, which have played a role in supplementing the representative system in existing policy processes.

Author contributions

All authors have conceptualization, writing the manuscript. All authors have read and agreed to the published version of the manuscript.

Data availability

The data that support the findings of this study are available from the first author [hongsoongoo@dtu.edu.vn] upon request.

Declarations

Conflicts of interest

The authors declare no conflict of interest.